FINDINGS FROM THE ALBERTA AND ONTARIO LONGITUDIL STUDIES OF GAMBLING Dr. Robert Williams Faculty of Health Sciences & Alberta Gambling Research Institute GARN Conference Stockholm, Sweden October 24, 2013
2
Utility of Cross-Sectional Surveys Cross-Sectional prevalence surveys are very good at: Establishing the population prevalence of gambling and each form of gambling Establishing the prevalence and number of people who are experiencing gambling problems Identifying factors correlated with gambling and problem gambling Identifying which subgroups have higher rates of problem gambling
Limitations of Prevalence Surveys Prevalence surveys are not very useful for: Identifying exactly how problem gambling (PG) develops at the individual level Correlates of problem gambling identified in a prevalence study tell you nothing about whether they caused PG, developed coincident with PG, or developed because of PG Determining the incidence of problem gambling Shedding light on the natural course and stability of problem gambling
Advantages of a Longitudinal Cohort A longitudinal cohort allows you to determine whether a stable prevalence rate (e.g., 2 out of 6 people) is due to continuity in the same individuals (left) or to an incidence rate equivalent to the remission rate (right). Y1 Y2 Y3 Y4 Y5 Y1 Y2 Y3 Y4 Y5
Advantages of a Longitudinal Cohort A longitudinal cohort also allows you to establish whether an increase in suicides from one year to another is related (left) or unrelated (right) to an increase in problem gambling. Y1 Y2 Y1 Y2 S S S S S S S S S S = Problem gambling S = Suicide S S 6
Advantages of a Longitudinal Cohort More generally, a longitudinal cohort allows you to potentially create a detailed etiological model of how gambling and problem gambling develops, continues, and remits Knowing the true risk and protective factors leading to PG is very important for the purposes of prevention Knowing the factors implicated in sustained recovery is very important for the purposes of treatment 7
LEISURE, LIFESTYLE, LIFECYCLE STUDY (LLLP) IN ALBERTA Funded by the Alberta Gambling Research Institute
LLLP Collaborators Nady el-guebaly, MD David Hodgins, Ph.D. Garry Smith, Ph.D. Don Schopflocher, Ph.D. David Casey, Ph.D. Shawn Currie, Ph.D. Bev West (Research Analyst) Ashley McInnes (Research Assistant) Dina Lavorato (Research Analyst) Jeanne Williams (Research Analyst)
LLLP Study 2006 2013 1808 Albertans 524 oversampled for at risk characteristics (>70 th percentile for gambling frequency or expenditure within age group) 5 initial age groups: 13-15; 18-20; 23-25; 43-45; 63-65 5 comprehensive assessments (2 4 hrs each; primarily selfadministered) with 18-20 month intervals between the start of each assessment 3.6% PG prevalence & 2.0% PG incidence (at baseline) 76.1% retention rate after 4 th assessment
LLLP Attrition Bias Males Non-Caucasians Single, less educated, attending school, More types of gambling, more time spent gambling (not frequency) Greater gambling problem severity
QUINTE LONGITUDIL STUDY (QLS) IN ONTARIO Funded by the Ontario Problem Gambling Research Centre
QLS Collaborators Bob Hann(co-PI) Patricia McLaughlin (Site Manager) Bev West (Research Analyst) Nick White (Research Assistant) Kate King (Research Assistant) Trevor Flexhaug(Technical Support) Dr. Don Schopflocher (Co-Investigator) Dr. Rob Wood (Co-Investigator) Dr. Earl Grinols (Consultant) Dr. Jan McMillen(Consultant)
Quinte Longitudinal Study Quinte Region in southeastern Ontario, Canada 2006 2011 4123 Ontario adults 1216 oversampled for at risk characteristics Initial age range: 17 90 5 comprehensive assessments (1.5 2.5 hrs each; all self-administered) with 12 months between start of each assessment 3.3% PG prevalence; 1.4% PG incidence (at baseline) 93.9% retention rate (because of the uniqueness of this accomplishment a 229 page manual has been developed)
Reasons for High Retention in QLS Convenient assessment options (office, at home, mail-in) Efficient and well-tested questionnaire Excellent staff Very engaged with cohort and project Attentive to needs of each participant Vigilant to all aspects of the project that would impact retention Proactive analysis of variables influencing re-recruitment and survey completion Efficient use of staff time Comprehensive and versatile Contact Database Incentives for both participants and staff Participants: $50, $30, $30, $35, $35 Staff: Significant $ bonuses for annual retention targets; increased independence in how they re-recruited the cohort
Approach to Developing an Etiological Model 1. Coordination of the QLS and LLLP analyses 2 separate analyses, but using same analytic approach Development of single etiological model that works for both data sets 2. Reducing # of IVs (from ~100) by identifying IVs that are either: Statistically correlated with PG status in the same year Predictors of PG in the subsequent year Predictors of IVs that predict subsequent PG IVs need to have one or more of these features in both data sets
Approach to Developing an Etiological Model 3. Using this subset of IVs in Multi Level Modelling to identify a) which variables best predict development of PG, and b) whether there are individual differences in growth curves suggestive of case groupings 4. Using Structural Equation Modelling to test our theoretical model of PG Note: only the univariate results available at this point
Dependent Variables: LLLP & QLS CPGI = Canadian Problem Gambling Index 5+ PPGM = Problem and Pathological Gambling Measure (Williams & Volberg, 2010; 2013) DSM-IV PY =NODS for QLS & Fisher DSM-IV-MR-Junior for LLLP DSM-IV-L = CIDI LLLP QLS CPGI (5+) X X PPGM DSM-IV-PY (3+) X (adolescents) DSM-IV-L Qualitative X Qualitative data = everyone with a calculated CPGI 3+ during an assessment period was asked what caused their gambling problems = 1,310 open-ended responses X X X
Stability of Problem Gambling Important to factor in measurement error when assessing the stability of PG over time Accuracy of self-report compromised by: short period of time participants given to answer the questions incomplete recall recency bias self-deception mood state social desirability genuine uncertainty about whether they meet the criteria we are asking about (guilt, financial problems, etc.)
Measurement Error 1 month test-retest reliability of CPGI Total Score: r =.78 (Canada 2001; n = 417; Ferris & Wynne, 2001) Total Score: r =.75 (Canada 2006/7; n = 328; Williams & Wood, 2006/7) Traditional 5 Categories: r =.61 (Canada 2006/7; n = 328; Williams & Wood, 2006/7) 5+ vs. 0-4: V=.54 (Canada 2006/7; n = 328; Williams & Wood, 2006/7)
Reliable Change Index (RCI) Difference in the person s score over 2 time periods divided by the standard error of difference between the 2 test scores: = 1 2 2( 1 1 ) 2 RCI scores provide a measure of the change in standardized units. Thus, a RCI of 1.96 or larger is needed for statistical significance at p <.05
Reliable Change Index: QLS & LLLP CPGI has average test-retest reliability of.765 Average SD of CPGI total scores over the 4 Time periods is 2.15 in LLLP and 1.86 in QLS over the 5 Time Periods Hence, a raw score increase or decrease of >3 at the subsequent time period is what is required for a statistically significant change
Stability of CPGI 5+ Problem Gambling in QLS using the RCI 2006-07 2007-08 2008-09 2009-10 2010-11 Red = PG; White = NPG; N = 219 (each row represents a case)
Stability of CPGI 5+ Problem Gambling in LLLP using the RCI Red = PG; White = NPG; N = 44 (each row represents a case)
Summary of PG Stability Findings Very good consistency in findings across the two data sets (QLS and LLLP) and between the two assessment instruments (PPGM and CPGI). Chronicity and Duration About half of problem gamblers are problem gamblers in only one time period. Chronic unremitting problem gambling is uncommon. Only one-third of problem gamblers are problem gamblers in 3 or more time periods Only one-quarter are problem gamblers in 4 or more time periods Only 10% have problem gambling in all 5 years. Risk of chronic problem gambling increases with each consecutive year of problem gambling status. Recovery The above results also mean that close to three-quarters of problem gamblers are observed to recover (no longer meet problem gambling criteria).
Summary of PG Stability Findings Relapse Of those that no longer meet problem gambling criteria, three-quarters do not relapse (at least during a 4-5 year time frame). Only a minority of people move in and out of problem gambling in a 4-5 year time period. Probability of relapse increases with increased prior duration of problem gambling. More Severe Forms CPGI 8+ and PPGM Pathological gamblers have surprisingly similar patterns of chronicity, recovery, and relapse to CPGI 5+ and PPGM Problem gamblers. Other Prevalence of problem gambling is decreasing over time, primarily related to fewer new cases each year. Longer time frames are needed to understand overall course of problem gambling.
Variables Best Predicting Future Problem Gambling (for the 134 PPGM PGs in QLS & the 43 CPGI 5+ PGs in LLLP) DEMOGRAPHICS QLS LLLP Age Gender Never Married Married Common-Law Separated Divorced Widowed
Variables Best Predicting Future Problem Gambling DEMOGRAPHICS Immigrant NonCaucasian Aboriginal African Asian (Eastern) Asian (Southern) European (Eastern) European (Western) Latin American Other Ethnicity QLS LLLP
Variables Best Predicting Future Problem Gambling RELIGION QLS LLLP Catholic Protestant Other Religion Athiest or Agnostic Religiosity Score
Variables Best Predicting Future Problem Gambling DEMOGRAPHICS No more than elementary school No more than technical college Completed college or university Unemployed Retired Homemaker Full-time Student On leave/strike Employed Part-time Employed Full-time QLS LLLP
Variables Best Predicting Future Problem Gambling INCOME & DEBT QLS LLLP Household (HH) Income Less than $30,000 HH Income Between $30,000 - $49,999 HH Income Between $50,000- $89,999 HH Income over $90,000 Average HH Debt
Variables Best Predicting Future Problem Gambling PHYSICAL HEALTH QLS LLLP Disability/Health Concern General Physical Health Rating Currently taking Prescription Medication
Variables Best Predicting Future Problem Gambling LIFETIME GAMBLING QLS LLLP Age first gambled Frequency of gambling prior to age 19 Big gambling win prior to 19 Big gambling win and loss priorto 19 Parents or siblings regular gamblers when growing up Parents or siblings gambledwith person when growing up Parents or siblings problem gamblers when growing up
Variables Best Predicting Future Problem Gambling GAMBLING QLS LLLP Attitude Toward Gambling Number of Different Gambling Formats Engaged in within Past Year
Variables Best Predicting Future Problem Gambling GAMBLING FREQUENCY IN PAST YEAR QLS LLLP ALL FORMS Lottery Purchase Instant Win Tickets Play Bingo EGM Play Casino Table Game Play Games of Skill for Money Sports Betting Horse Race Betting High Risk Stocks/Options/Futures
Variables Best Predicting Future Problem Gambling PAST YEAR SPENDING ON GAMBLING QLS LLLP OVERALL spending per month on gambling Spending per month on EGMs Largest amount won in single day Largest amount lost in single day Frequency of ATM use in gambling venues
Variables Best Predicting Future Problem Gambling GAMBLING MOTIVATION & FALLACIES QLS LLLP Gamble for fun/excitement/entertainment Gamble to win money Gamble to escape/distraction Gamble to socialize Gamble to support worthy causes Gamble to feel good about oneself Gambling Fallacies
Variables Best Predicting Future Problem Gambling GAMBLING CONTEXT QLS LLLP Membership in Gambling Rewards Program Gamble alone vs with friends Drink alcohol when gambling Use tobacco when gambling Use [street] drugs when gambling Gambling on Internet
Variables Best Predicting Future Problem Gambling GAMBLING SOCIAL EXPOSURE QLS LLLP Number of close friends/family that are regular gamblers Number of close friends/family that have gambling problems Other adults in household with gambling problems Opportunities to gamble at workplace (or school) Exposed to prevention or awareness campaign
Variables Best Predicting Future Problem Gambling PROXIMITY TO EGM VENUES QLS LLLP Driving time to nearest venue with EGMs Actual distance to nearest venue with EGMs Casinos/Racinos within 5km driving distance
Variables Best Predicting Future Problem Gambling GAMBLING AS LEISURE ACTIVITY QLS LLLP Gambling as one of 5 favouriteleisure activities Gambling as favorite leisure activity
Variables Best Predicting Future Problem Gambling PERSOLITY(NEO-PI) & COPING SKILLS QLS LLLP Neuroticism Depression Vulnerability Impulsivity Extraversion Excitement Seeking Agreeableness Conscientiousness Openness Coping Skills
Variables Best Predicting Future Problem Gambling TRAUMA QLS LLLP Abused as a Child Past Year Post-Traumatic Stress Disorder (CIDI)
Variables Best Predicting Future Problem Gambling STRESS & WELL-BEING QLS LLLP Number of Stressful Events in Past Year Average Stress Level Average Happiness Average Life Satisfaction Personal Wellness Index Score
Variables Best Predicting Future Problem Gambling PERSOL VALUES Money is Most Important Power is Most Important Fame is Most Important Friendship is Most Important None of the Above is Most Important Wealth is a Good Indicator of Success QLS LLLP
Variables Best Predicting Future Problem Gambling LIFETIME ADDICTION & MENTAL HEALTH QLS LLLP Lifetime history of drugs/alcohol addiction Lifetime history of addiction to other behaviours (not including gambling) Lifetime history of mental health problems (prior to past year) Parents or siblings with history of other addictions (not including gambling) Parentsor siblings with history of mental health problems
Variables Best Predicting Future Problem Gambling MENTAL HEALTH PROBLEMS IN PAST YEAR QLS LLLP ANY MENTAL HEALTH PROBLEM Major Depression Manic Episode Generalized Anxiety Panic Attacks Obsessive Compulsive Disorder Eating Disorder Schizophrenia and/or Delusional Disorder Specific Phobias ADHD
Variables Best Predicting Future Problem Gambling PAST YEAR ADDICTION QLS LLLP Substance Abuse (ASSIST for QLS) Substance Dependence (ASSIST for QLS) Any Other Addiction (sex, video games, exercise, etc.) Other Adultsin Household with Substance Abuse
Variables Best Predicting Future Problem Gambling SOCIAL FUNCTIONING & SUPPORT QLS LLLP Marital Satisfaction (if applicable) Overall Family Functioning Social Support (PAI) Community Involvement Job Satisfaction (if applicable) Job Stress (if applicable)
Variables Best Predicting Future Problem Gambling SEXUAL ORIENTATION QLS LLLP Heterosexual Bisexual, Homosexual, or Prefer not to say
Variables Best Predicting Future Problem Gambling ILLEGAL ACTIVITY AND ANTISOCIALITY QLS LLLP Number of Illegal Behaviours in Lifetime Number of Illegal Behaviours in Past Year Antisociality(PAI)
Variables Best Predicting Future Problem Gambling INTELLIGENCE & FRONTAL LOBE FUNCTIONING QLS LLLP Overall Intelligence (SB-IV or WAIS-R) Below Average Intellect Wisconsin Card Sorting Test
MOST IMPORTANT & ROBUST VARIABLES BEST PREDICTING FUTURE PROBLEM GAMBLING FROM THE QLS & LLLP STUDIES Demographics NonCaucasian(particularly East Asian) Gambling Early big win Family member(s) regular gamblers &/or problem gamblers Gambling with family prior to 19 Number of formats engaged in Overall Frequency of play Overall Gambling expenditure Membership in gambling rewards program Gambling as a favoured leisure activity Current friends/family with gambling problems
MOST IMPORTANT & ROBUST VARIABLES BEST PREDICTING FUTURE PROBLEM GAMBLING FROM THE QLS & LLLP STUDIES Mental Health Lifetime history of addiction Presence of any mental health disorder Substance abuse or dependence Presence of any behavioural addiction Other Illegal activities in past year Below average intellect
RELEVANCE TO MODEL DEVELOPMENT Big win; positive attitude toward gambling; early gambling; gambling among and within family and friends; lower education; lower intelligence; personality (excitement seeking; low openness); cultural tradition; younger age Stress; biological predisposition to mental health problems; upbringing; trauma Heavy Gambling Involvement Mental Health Problems Problem Gambling